Benoit LOGER Thesis Defense

Abstract: With the growing complexity of supply chains, the use of automated decision-support tools is becoming necessary to apprehend the multiple sources of uncertainty likely to impact them, while guaranteeing a high level of performance. To meet these objectives, managers are increasingly turning to approaches capable of improving the resilience of supply chains by proposing robust solutions in the face of hazards, to guarantee both quality of service and control of costs linked to the production, storage and transportation of goods. At a time when data collection and analysis are playing an increasingly important role in corporate strategy, the proper use of this information to more accurately characterize these uncertainties and their impact on operations is becoming a major challenge for optimizing modern production and distribution systems. This thesis addresses these new challenges by developing different mathematical optimization methods based on the use of historical data, with the aim of proposing robust solutions to several procurement and production planning problems. Numerical experiments on a wide range of applications are used to compare these new techniques with several other classical approaches in the literature, and to validate their practical relevance. The results obtained demonstrate the value of these contributions, which offer comparable average performances while reducing their variability in an uncertain context. In particular, the solutions remain satisfactory when confronted with extreme scenarios, whose probability of occurrence is low. Last but not least, the resolution times of the procedures developed remain competitive, making them suitable for industrial-scale applications.